Thank you, we leave it as it is, so one last doubt about parameterization.

In the example I read before, the formula of Cox’s model started with the term “1+”.

Therefore, I used this syntax:

fit <- brm_multiple(days | cens(1 - event) ~ 1 + treatment + x2 + x3, data = imputed, prior=prior, family = brmsfamily(“cox”))

Since the parameterization initially changed, as I read, I wanted to ask if (1) the term (“1+”) is still necessary inside the model formula, (2) if the interpretation is finally the usual one of the Cox model, i.e. negative coefficients (or hazard ratios <1 in its exponentiated version) are associated with less risk of the event.

Thanks.

# Brms survival model with interactions between a binary variable and a spline function

+1 (adding an intercept) is equivalent to leaving it out. so it doesn’t matter if you add it or now. The current version of brms uses the standard cox parameterization that is used also by other packages, i.e. negative values indicate less risk and longer surival times.

1 Like

Thanks a lot for the response !